tests/testthat/test_arma_utils.R

In baruuum/btoolbox: random toolbox

context("distance/cosine similarity measures")

test_that("measures work properly", {
  
    r_dist = function(x, y, what = c("cos_sim", "pdist"), p = NULL) {
      
        what = match.arg(what)
        if (is.null(p)) p = 2
        
        res = matrix(NA, nrow = NROW(x), ncol = NROW(y))
      
        for (i in 1:NROW(x)) {
            for (j in 1:NROW(y)) {
                
                rx = x[i, ]
                ry = y[j, ]
                
                res[i, j] = if (what == "cos_sim") {
                        sum(rx * ry) / sqrt(sum(rx^2)) / sqrt(sum(ry^2))
                    } else {
                        
                        if (p == 1) {
                          
                            sum(abs(rx - ry))
                          
                        } else {
                          
                            sum((rx - ry)^p)^(1/p)
                          
                        }
                    }
            }
        
        }      
        
        return(res)
        
    }
    
    for (rr in 1:5) {
      
        n = 20
        k = 4
        A = matrix(runif(n * k) + 1, ncol = k)
        B = matrix(runif((n - 3) * k) + 1, ncol = k)
        
        # cosine similarity
        res1 = row_cos_sim(A, B)
        res2 = frow_cos_sim(A, B)
        r_res = r_dist(A, B, "cos_sim")
        expect_equal(res1, res2)
        expect_equal(res1, r_res)
        
        # Euclidean distance
        res3 = row_pdist(A, B, 2L)
        res4 = frow_euc_dist(A, B)
        res5 = r_dist(A, B, "pdist", p = 2)
        expect_equal(res3, res4)
        expect_equal(res3, res5)
        
        # Manhattan distance
        res6 = row_pdist(A, B, 1)
        res7 = r_dist(A, B, "pdist", p = 1)
        expect_equal(res6, res7)
        
        # Minkowski distance
        k = max(rpois(1, 2), 1)
        res8 = row_pdist(A, B, k)
        res9 = r_dist(A, B, "pdist", p = k)
        # skip if r-code results in NaNs
        if (sum(is.na(res9)) == 0)
            expect_equal(res8, res9)    
        
        # compare with built-in distance function
        res10 = as.matrix(stats::dist(A))
        res11 = row_pdist(A, A, 2)
        expect_equivalent(res10, res11)
        
        res12 = as.matrix(stats::dist(A, method = "minkowski", p = k))
        res13 = row_pdist(A, A, k)
        expect_equivalent(res12, res13)
        
        # check double-centering
        A_c = double_center(A)
        expect_equal(rowSums(A_c), rep(0, nrow(A)))
        expect_equal(colSums(A_c), rep(0, ncol(A)))
        
    }
    
})